# Test 3a: Cohort-Specific Pre/Post Z₁ Attenuation

*If composition effects drive attenuation, the pattern should be cohort-specific, not universal.*

| Cohort | Period | Z₁ | SE | p | R² | N | Countries |
|:---|:---|---:|---:|---:|---:|---:|---:|
| Early (<2000) | Post | 163.866 | (108.797) | 0.135 | -0.300 | 124 | 5 |
| Late (≥2000) | Pre | 199.403* | (109.876) | 0.073 | -0.348 | 104 | 12 |
| Late (≥2000) | Post | -1.382 | (83.141) | 0.987 | 0.249 | 186 | 13 |
| All openers | Pre | 177.847* | (100.328) | 0.079 | -0.088 | 110 | 15 |
| All openers | Post | 62.222 | (72.695) | 0.393 | -0.130 | 310 | 18 |

## Attenuation Ratios

- **Late (≥2000)**: Z₁ pre=199.403, post=-1.382, post/pre ratio=-0.01
- **All openers**: Z₁ pre=177.847, post=62.222, post/pre ratio=0.35

*Controls: fiscal_bal_gdp, nfa_gdp_lag, trade_openness, log_rel_opw.*
*Transition economy openers only. PanelGLS with entity and time effects.*